paper-with-me

Papers

Towards Minimal Targeted Updates of Language Models with Targeted Negative Training

2024-06-19 · Lily H. Zhang, Rajesh Ranganath, Arya Tafvizi

Generative models of language exhibit impressive capabilities but still place non-negligible probability mass over undesirable outputs. In this work, we address the task of updating a model to avoid unwanted outputs while minimally changing model behavior otherwise, a challenge we refer to as a minimal targeted update. We first formalize the notion of a minimal targeted update and propose a method to achieve such updates using negative examples from a model's generations. Our proposed Targeted Negative Training (TNT) results in updates that keep the new distribution close to the original, unlike existing losses for negative signal which push down probability but do not control what the updated distribution will be. In experiments, we demonstrate that TNT yields a better trade-off between reducing unwanted behavior and maintaining model generation behavior than baselines, paving the way towards a modeling paradigm based on iterative training updates that constrain models from generating undesirable outputs while preserving their impressive capabilities.

📄 PDF Abstract BibTeX arXiv:2406.13660

Code (1)

google/t5patches 공식 구현 jax

Methods 이 논문이 사용한 방법론

TNT Transformer is a type of self-attention-based neural networks originally applied for NLP tasks. Recently, pure transformer-based…

Similar Papers 제목 키워드 기반

Constructive Circuit Amplification: Improving Math Reasoning in LLMs via Targeted Sub-Network Updates

2025-12-18 · Nikhil Prakash, Donghao Ren, Dominik Moritz, Yannick Assogba arxiv

Prior studies investigating the internal workings of LLMs have uncovered sparse subnetworks, often referred to as circuits, that are responsible for performing specific tasks. Additionally, it has been shown that model p…

Mathematical Reasoning

FL-Defender: Combating Targeted Attacks in Federated Learning

2022-07-02 · Najeeb Jebreel, Josep Domingo-Ferrer

Federated learning (FL) enables learning a global machine learning model from local data distributed among a set of participating workers. This makes it possible i) to train more accurate models due to learning from rich…

Federated Learning

Data Poisoning Attacks Against Federated Learning Systems

2020-07-16 · Vale Tolpegin, Stacey Truex, Mehmet Emre Gursoy, Ling Liu

Federated learning (FL) is an emerging paradigm for distributed training of large-scale deep neural networks in which participants' data remains on their own devices with only model updates being shared with a central se…

Data PoisoningFederated Learning

No Free Lunch in Language Model Bias Mitigation? Targeted Bias Reduction Can Exacerbate Unmitigated LLM Biases

2025-11-23 · Shireen Chand, Faith Baca, Emilio Ferrara arxiv

Large Language Models (LLMs) inherit societal biases from their training data, potentially leading to harmful or unfair outputs. While various techniques aim to mitigate these biases, their effects are often evaluated on…

Parameter-Efficient Token Embedding Editing for Clinical Class-Level Unlearning

2026-03-11 · Iyad Ait Hou, Shrenik Borad, Harsh Sharma, Pooja Srinivasan 외 arxiv

Machine unlearning is increasingly important for clinical language models, where privacy regulations and institutional policies may require removing sensitive information from deployed systems without retraining from scr…